Ankur: Bangla online character recognition

Bangla alphabet consists of 50 characters. Most of these characters have twists and turns that is very complicated to trace down in real time for an online OCR (Optical Character Recognizer). The main challenge of making a reliable Bangla OCR is to generate analysable features of these curvaceous characters in a real time environment. For an online Bangla OCR, fuzzy logic could be one the most efficient way as it has very low computational requirement. This ensures highest efficiency with economical use of the processing power. The most important part of creating an online OCR is establishing a reliable fuzzy rule-base that would describe the characters to be recognized. However, for Bangla this job is quite cumbersome due to the curvaceous nature of the characters and, not to mention, variety of handwritings of different people. This paper aims to describe a way that can be used to generate a fuzzy-feature database that describes Bangla characters written in different handwritings. It also suggests a system for recognizing any given character with respect to the linguistic variable extracted from the fuzzy-feature database.